|Home | About | Journals | Submit | Contact Us | Français|
Trinucleotide repeat (TNR) expansions are the underlying cause of more than forty neurodegenerative and neuromuscular diseases, including myotonic dystrophy and Huntington’s disease. Although genetic evidence has attributed the cause of these diseases to errors in DNA replication and/or repair, clear molecular mechanisms have not been described. We have focused on the role of the mismatch repair complex Msh2-Msh3 in promoting TNR expansions. We demonstrate that Msh2-Msh3 promotes CTG and CAG repeat expansions in vivo in Saccharomyces cerevisiae. We further provide biochemical evidence that Msh2-Msh3 directly interferes with normal Okazaki fragment processing by flap endonuclease1 (Rad27) and DNA Ligase I (Cdc9) in the presence of TNR sequences, thereby producing small, incremental expansion events. We believe that this is the first mechanistic evidence showing the interplay of replication and repair proteins in the expansion of sequences during lagging strand DNA replication.
Errors in DNA replication are hypothesized to cause trinucleotide repeat (TNR) sequences to expand (McMurray, 2010), specifically during lagging strand replication (Ireland et al., 2000; Schweitzer and Livingston, 1998). Lagging strand maturation requires Okazaki fragments to be processed and joined into a continuous strand. DNA polymerase δ (pol δ) is responsible for the extension of Okazaki fragments and initiates their processing by displacing the 5’ end of one fragment while extending the 3’ end of the preceding fragment (Balakrishnan and Bambara, 2011). The newly displaced DNA forms a 5’ flap structure, which is cleaved by flap endonuclease1 (Rad27) (Bambara et al., 1997; Lieber, 1997). This creates a nick which is subsequently sealed by DNA Ligase I (Cdc9) (Balakrishnan and Bambara, 2011).
Genetic studies in yeast suggest that both Rad27 and Cdc9 prevent TNR repeat expansions, as the down-regulation of either protein leads to a higher frequency of expansions (Ireland et al., 2000; Schweitzer and Livingston, 1998). The mismatch repair (MMR) complex Msh2-Msh3 has also been implicated in TNR expansion; the knockdown of either subunit leads to a decrease in expansion rate of CTG and CAG repeats in a mouse model for Huntington’s disease (Manley et al., 1999; Owen et al., 2005). Similarly, a lower rate of expansion was also reported in Msh3-deficient DM1 transgenic mice, a model for myotonic dystrophy type I (Foiry et al., 2006; van den Broek et al., 2002). Eukaryotic MMR is initiated by the recognition of mispaired sequences, either those resulting in small loops (1–13 nucleotides in length) or single nucleotide loops and mispairs, by the Msh2-Msh3 or Msh2-Msh6 repair complex respectively (Kolodner and Marsischky, 1999). Since the main function of the mismatch proteins is to maintain genome integrity, the role of Msh2-Msh3 in promoting repeat expansions has been an area of intense interest in the past few years.
Here we demonstrate for the first time that Msh2-Msh3 contributes to CTG and CAG repeat expansion in S. cerevisiae. We further present mechanistic details of how Msh2-Msh3 can promote TNR expansions by altering the activity of Rad27 and Cdc9 on DNA intermediates formed during Okazaki fragment processing (OFP). These details were revealed by the dynamic nature of the DNA substrates we employed, which, in contrast to previous studies, were free to adopt various TNR structure conformations. Using this unique system, we observed a bias toward incremental TNR expansion in the presence of Msh2-Msh3.
Msh2-Msh3 has been implicated in repeat expansion in mice, and since we intended to examine this phenomenon in yeast, we first questioned whether the Msh complex is required for expansions in S. cerevisiae. We used a reporter system developed to detect lagging strand expansions in S. cerevisiae (Dixon et al., 2004; Miret et al., 1998). The sequences (CTG)25, (CAG)25, or scrambled (C,A/T,G)25 were positioned on the lagging strand, fused to a URA3 reporter gene (Fig S1). Expansions of five or more repeats cause inactivation of the URA3 gene and resistance to 5-FOA. Since msh2Δ has essentially the same expansion rate as wild-type (Miret et al., 1998), we chose to examine the effect of msh3Δ Δor msh6Δ, whose gene products interact with Msh2 to form two distinct heterodimeric complexes. Cells harboring an msh3Δ mutation showed a significant decrease in expansion rates for both CTG (~10–20 fold) and CAG (~5 fold) sequence repeats on the lagging daughter strand, as compared to the wild type strain (Table 1; p<0.0001 for each). This result is specific for the Msh2-Msh3 complex, as msh6Δ cells did not show a decrease but rather a significant increase in expansion rate, at least for the CTG repeat (~5 fold) (Table 1; p=0.0091). Furthermore, the scrambled sequences did not show any significant difference in expansion rate between wild type and mutant strains. These data suggest that the absence of an expansion phenotype in msh2Δ cells could derive from opposing effects of Msh2-Msh3 and Msh2-Msh6 on lagging strand synthesis. Our results show, for the first time, that similar to mice, Msh2-Msh3 promotes TNR expansions in S. cerevisiae, solidifying the yeast system as a relevant model to characterize the mechanism of expansion.
Msh2-Msh3 has affinity for DNA substrates containing double-strand (ds)/single-strand (ss) junctions. Specifically, Msh2-Msh3 has a preference for binding junctions with 3’ ssDNA, consistent with its role in mismatch repair and double strand break repair, and also binds 5’ flap intermediates that form during Okazaki fragment processing (Surtees and Alani, 2006). To determine whether Msh2-Msh3 binds preferentially to the base of a 5’ flap, we performed 1, 10-phenanthroline-copper (OP-Cu) footprinting on duplex DNA containing an 18 nt unannealed region, simulating an 18 nt, 5’ flap intermediate. Msh2-Msh3 bound to the base of a 5’ flap (Fig 1a and S2), with a pattern that is similar to that observed with the 3’ flap (Surtees and Alani, 2006). This is also the site that Rad27 initially binds before cleavage (Gloor et al., 2010; Tsutakawa et al., 2011). To determine whether Msh2-Msh3 can influence Rad27 flap processing, we performed cleavage assays using 5’ flap substrates with varying flap lengths. A titration of Msh2-Msh3 in the presence of Rad27 resulted in a significant decrease (3–5 fold) of cleavage products for all flap lengths tested (Fig 1b). To verify the binding specificities of Rad27 and Msh2-Msh3 on a 5’ flap substrate, we performed gel shift assays using a 10 nt and 20 nt flap substrate. After pre-binding Rad27 to the substrates (Fig 1c, lanes 2 and 9) Msh2-Msh3 was titrated into the reactions (Fig 1c, lanes 3–7 and 10–14). Msh2-Msh3 competed with Rad27 for binding to flap substrates, indicated by the disappearance of the Rad27-bound band upon titration of Msh2-Msh3. In combination with the footprinting data, the cleavage and gel shift assay results suggest that Msh2-Msh3 can inhibit Rad27 flap processing by competing for binding to the base of a 5’ flap intermediate (Fig 1b).
Another preferred structure for Msh2-Msh3 binding is a DNA bubble (Surtees and Alani, 2006), a structure known to form during polymerase pausing and subsequent DNA misalignment (Liu and Bambara, 2003). It was previously proposed that the ligation of Okazaki fragments containing such structures can lead to sequence expansion (Wells, 1996). To determine whether the activity of Cdc9 is affected in the presence Msh2-Msh3, we designed substrates containing a 12 nt unannealed bubble in the downstream primer. These substrates also contained a nick at a distance of 2, 4, or 6 nt of complementary DNA from the 5’ end of the bubble. The substrate with the 2 nt annealed region could not be ligated into an expanded product (Fig 1d, lane 2), as it may not form a stable bubble structure because of DNA breathing (Liu and Bambara, 2003). Even though the 4 nt and 6 nt substrates were ligated, the latter was ligated more efficiently (Fig 1d, lanes 9 and 16). In the presence of Msh2-Msh3, the ligation efficiency of the 4 nt substrate was reduced (Fig1d, lanes 10–14). Ligation of the 6 nt substrate was unaffected. OP-Cu footrprinting of Msh2-Msh3 bound to a 12 nt bubble substrate revealed that Msh2- Msh3 interacts with and distorts DNA up to 4 nt from the 5’ end of the bubble (Fig 1a and S2). The footprinting analysis correlates with the ligation assay results, since the position of the nick influences Cdc9 activity when within but not outside of the Msh2-Msh3 binding region (4 nt or 6 nt from the bubble respectively) (Fig 1d). Additionally, even though the footprinting data show that Msh2-Msh3 distorts DNA structure, it did not promote the ligation of a DNA bubble into an expanded sequence (Fig 1d, lanes 3–7). Thus, our results suggest that for substrates that cannot be efficiently ligated (4 nt substrate), Msh2-Msh3 further inhibits their ligation by binding to the nicked sites and blocking Cdc9 activity.
The presence of Msh2-Msh3 can significantly influence the activity of both Rad27 and Cdc9 on their cognate substrates; therefore, we asked whether it could affect their combined activity on TNR-containing replication intermediates. During replication at TNR tracts, the polymerase has the potential to pause and dissociate from the DNA (Kang et al., 1995; Viguera et al., 2001). Such events allow for the newly synthesized strand and displaced flap to compete for annealing to the template (Liu and Bambara, 2003). To simulate such intermediates, we designed equilibrating flap substrates in which the downstream and upstream primers overlap and compete for template annealing over a 30 nt region. Therefore, the substrate can equilibrate into different conformations, with either 5’ and/or 3’ flaps of various sizes, depending on how the primers anneal to the template. The equilibrating sequences consisted of either ten CTG repeats or 30 nt of a randomly generated sequence, which was predicted not to form any structure (Fig 2a). Similar to previous reports showing that such large flaps cannot be effectively ligated (Veeraraghavan et al., 2003), we did not observe the formation of a fully expanded product (104 bp) in the presence of Cdc9 (Fig 2a, lane 3). Incubation of the substrate with Rad27 and Cdc9 resulted in the majority of products being correctly processed (74 bp) (Fig 2a, lane 4), since Rad27 effectively cleaved at the base of the flap, creating a nick for Cdc9 to seal. Interestingly, a minority of various length expansion products was observed only with the substrate containing CTG repeats (Fig 2a, lane 4) and not the randomly generated sequence (Fig 2a, lane 12). Significantly, the titration of Msh2-Msh3 into reactions containing both Rad27 and Cdc9 resulted in the decrease of the correctly processed product and the preferential increase of only smaller expansion products for TNR substrates, corresponding to one or two repeats (Fig 2a, lanes 5–8, Fig 2b). This increased the ratio of partially expanded to correctly processed products ~ 4.3 fold as compared to the ratio produced without Msh2-Msh3.
It was previously reported that Rad27 cleavage of such intermediates containing TNRs generates multiple cleavage products because the two competing sequences interact with the template at various positions and equilibrate into different conformations (Liu et al., 2009; Veeraraghavan et al., 2003). This produces various length 5’ flaps that can be cleaved and ligated to the upstream primer to create expanded products (Fig 2b). In order to determine whether these expanded products could also be a result of Rad27 cleavage upstream of the flap base, we examined substrates containing only a static 5’ flap consisting of either ten CTG, four CTG, or four CAG repeats. Alternate Rad27 cleavage patterns could be observed (Fig S3a), presumably responding to secondary structure formation within the 5’ flap (Vallur and Maizels, 2010). Even though Rad27 cleaved at multiple sites upstream of the base (Fig S3a), only a single 2 nt expansion product was observed with the addition of Cdc9 (Fig S3b and S3c), corresponding to a single flap conformation that can be processed into an expanded sequence for this static substrate. Interestingly, the presence of Msh2-Msh3 in the reactions promoted the accumulation of the expanded product and decreased the formation of the correctly processed product (Fig S3b and S3c). These results indicate that cleavage and ligation patterns of TNR-containing 5’ flaps can be altered by the formation of structure within the flap, an effect that is enhanced by the interaction with Msh2-Msh3.
Our findings with a large equilibrating flap indicate that Msh2-Msh3 favors the creation of small expansions instead of large. We, therefore, investigated the effect of Msh2-Msh3 on a smaller equilibrating flap substrate that can efficiently ligate into an expanded sequence. This substrate was designed to contain small equilibrating 9 nt flaps consisting of three CTG repeats, that would compete for annealing to the template. Additionally, the downstream primer contained seven more CTG repeats that could fully anneal to the template without competition, in order to enhance the possible structural conformations the substrate could adopt. Incubation with Rad27 resulted in a predominant population of cleavage products that correspond to cleavage at the base of the flap (39 bp) (Fig 2c, lane 2). A fully expanded product was also observed upon incubation with Cdc9 (75 bp) (Fig 2c, lane 3). The combined addition of Rad27 and Cdc9 produced both a correctly processed (66 bp) and fully expanded product (Fig 2c, lane 4). Upon titration of Msh2-Msh3, however, smaller expansion products (69 and 72 bp) were predominant over the larger fully expanded product (Fig 2c, lanes 5–9). These products are not observed with substrates containing short equilibrating flaps consisting of randomly generated sequences (data not shown). Specifically, Msh2-Msh3 causes up to a ~3.3 fold increase in the accumulation of smaller expansion products (Fig 2c, lane 9). These data are all consistent with Msh2-Msh3 binding and stabilizing small TNR structures that form within the flaps.
Our results suggest that Msh2-Msh3 promotes the accumulation of expanded products not only by inhibiting Rad27 cleavage at the base of 5’ flaps, but also by stabilizing small loop structures (Lang et al., 2011) formed during the equilibrating conformations of TNR OFP intermediates (Fig 2b, Fig 3). Since our analysis of static TNR 5’ flaps indicated that they are structured, it is likely that a 3’ flap of the same sequence would form structure as well. It is, thus, possible that for equilibrating TNR flaps the Msh complex not only inhibits the correct processing of 5’ flaps, but also binds and stabilizes small loops formed within the 3’ flap, consistent with previous reports showing that Msh2-Msh3 has a binding preference for 3’ flap over 5’ flap intermediates (Surtees and Alani, 2006). This would allow for the downstream 5’ flap to anneal farther along the template and vary in size (Fig 2d, Fig 3). The cleavage and ligation of such shorter flaps could lead to the creation of an expanded sequence.
Furthermore, our data reveal that the Msh complex also inhibits cleavage of non-TNR flap intermediates. Previous studies have suggested that although Msh2-Msh3 can efficiently dissociate from non-TNR DNA structures, it cannot as readily dissociate from TNR structured DNA and remains in a ‘stuck’ bound conformation (Lang et al., 2011; Owen et al., 2005). Hence, even though we observe Rad27 cleavage inhibition on non-TNR flaps by Msh2-Msh3 in vitro, cellular conditions may regulate the dissociation of the Msh complex from non-TNR repeat structures. The stuck conformation of Msh2-Msh3, however, would allow the protein complex to interfere more effectively with the activity of the replication proteins only on TNR intermediates. Even though Msh2-Msh3 DNA dissociation is dependant on ATP hydrolysis (Gupta et al., 2011; Surtees and Alani, 2006), it has been shown that it remains stuck on TNR sequences even in the presence of ATP (Lang et al., 2011; Owen et al., 2005); we, therefore, repeated our experiments with ATP. Our results show that expanded products occur at comparable levels to those reported in Figure 2 (data not shown).
The proposal of a stuck conformation of Msh2-Msh3 bound to TNR structures (Lang et al., 2011; Owen et al., 2005) has focused current research on the ability of the complex to initiate or block the repair of these structures. Interestingly, studies have indicated that Msh2-Msh3 does not interfere with the repair of TNR structures in vitro and can even promote repair of small loops (Panigrahi et al., 2005; Panigrahi et al., 2010; Tian et al., 2009), raising questions about the mechanistic role of Msh2-Msh3 in DNA expansion. These studies, however, were performed in vitro using substrates having a pre-formed TNR structure. In the current study, we have instead used equilibrating substrates, which have the propensity for forming TNR structures of varying size. Our results using this dynamic system supports a model in which the role of Msh2-Msh3 in TNR expansion is in shifting the equilibrium of flaps into many structured conformations, which would affect processing by RAD27 and Cdc9. Even though we observe this effect on equilibrating structures of replication intermediates, it need not necessarily inhibit repair at all TNR structures.
Our current results indicate a mechanism by which Msh2-Msh3 contributes to small length TNR sequence expansions during OFP (Fig 3). Importantly, Msh2-Msh3 promotes small length expansions, leading to the incorporation of one or two repeats. We believe that this model for expansion initiation supports the observation that the progression of expansions per replication cycle is very slow (McMurray, 2010). This is in contrast to both non-dividing neuronal cells and undifferentiated germ cells, in which expansions events involve production of greater length sequences (McMurray, 2010). Evidently, additional mechanisms exist to expand TNRs, including errors made during DNA base excision repair (BER) (Liu and Wilson, 2012). Additionally, our footprinting analysis of the 5’ flap substrate suggests that Msh2-Msh3 has access to lagging strand replication intermediates, perhaps to help localize mismatch repair to the lagging strand. Thus, while MMR may be active on both replicating strands our results suggest that the Msh complex can interact directly with the lagging strand during processing, consistent with reports suggesting that mismatch repair occurs more frequently on this strand (Pavlov et al., 2003). Although the role of Msh2-Msh3 in expansions is counterintuitive, our model suggests that its involvement is an undesirable side-effect of its main function as a repair complex, in binding to small loops containing mismatches (Kolodner and Marsischky, 1999). We expect that our findings on the Msh complex role in TNR sequence expansions will contribute to a better mechanistic understanding of how expansions result in neuromuscular diseases.
S. cerevisiae Cdc9 (Cdc9p) and Msh2-Msh3 were overexpressed in S. cerevisiae and purified as described previously (Ayyagari et al., 2003; Lee et al., 2007; Surtees and Alani, 2006). S. cerevisiae Rad27 (Rad27p) was overexpressed in E. coli and purified as described previously (Kao et al., 2002).
TNR substrates were integrated into WT, msh3Δ and msh6Δ strains, as described previously (Dixon et al., 2004; Miret et al., 1998). Four different substrates were used and integrated into the three genetic backgrounds, with the following sequence on the lagging daughter strand: 1) a (CTG)25 repeat (pBL69); 2) a (CAG)25 repeat (pBL70); 3) a scrambled (C,T,G)25 repeat (pBL138) and 4) a scrambled (C,A,G)25 repeat (pBL139). Each repeat tract was cloned into the regulatory region controlling expression of the URA3 reporter gene. When the distance between the TATA box and the initiator ATG for the URA3 gene is increased beyond 29 repeats, URA3 is no longer expressed, making the cells resistant to 5-FOA. Colonies were then counted and expansion rates calculated as described by Drake (Drake, 1991). The 95% confidence intervals were determined from tables of confidence intervals for the median (Dixon, 1969; Nair, 1940). p values were determined by Mann-Whitney rank analysis. Also, see supplementary experimental procedures.
All reactions were performed in 20µl containing 1X MSH buffer (30mM HEPES, 40mM KCl, 4mM MgCl2, 0.01% NP-40, 0.5% Inositol, 0.1mg/ml BSA, 1mM DTT, and 5% glycerol). All reactions were mixed with 0.25nM DNA substrate on ice and incubated at 37°C for 10min. Fold inhibition of Rad27 cleavage was calculated by dividing the percent cleaved in the presence of Msh2-Msh3 by percent without (Balakrishnan et al., 2010). Fold inhibition of Cdc9 activity was calculated by dividing the percent ligated in the presence of Msh2-Msh3 by the percent without (Balakrishnan et al., 2010). Expansions were calculated as a percent of expanded product/percent of correctly processed product (Balakrishnan et al., 2010). Fold expansion was calculated by dividing the percent expansion in the presence of Msh2-Msh3 by the percent without. Also, see supplementary experimental procedures.
We thank Dr. Robert Lahue for the plasmids to generate trinucleotide repeat expansion substrates, Dr. Eric Alani and Dr. Mark Sutton for a careful reading of the manuscript, and Dr. Peter Burgers for sharing purified S. cerevisiae DNA Ligase I. Lastly, we would like to thank members of the Bambara and Surtees laboratories for insightful discussions. This work was supported by National Institutes of Health Grants GM098328-01A1 (to L.B.), GM087549 (to J.A.S.), and GM024441 (to R.A.B.).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.